Upgrade to Pro — share decks privately, control downloads, hide ads and more …

Beating the competition with Cognitive Commerce

Meanbee
April 05, 2017

Beating the competition with Cognitive Commerce

Artificial Intelligence, and particularly machine learning, is the talk of the tech industry and how it’s going to revolutionize every sector. So what is it? What do you need to know? And how can you use it as a tool to get an edge? This session will look at some of the recent developments and use cases, and how they can be used to differentiate and excel in eCommerce. Topics will range from intelligent search, to customer service channels of Facebook Messenger and Amazon Alexa, and to finding swimming pools in photos with IBM Watson.

Meanbee

April 05, 2017
Tweet

More Decks by Meanbee

Other Decks in Technology

Transcript

  1. Meanbee • UK eCommerce Agency • Specialized in Magento •

    Technology First • Client revenues average $2-10 million
  2. Machine Learning • Subset of AI • Ability to spot

    correlations between properties and effects • Supervised • Unsupervised • Reinforcement
  3. Supervised Learning • This is the one we’re quite good

    • Provide example inputs and responses • Learn correlations between them
  4. Why now? Computers are fast enough to train on large

    datasets in a reasonable time with the same or better accuracy than a person
  5. Cognitive Commerce • It’s about being smarter: • Being smarter

    in how we interact with customers • Being smarter in how we manage our store • Being smarter in how we spend our time
  6. Automation • Tasks that we don’t need to be doing

    • Tasks that weren’t previously feasible
  7. What can be automated? • A good heuristic from Andrew

    Ng: “” Anything that a human can do with less than one second of thought
  8. Automate or be automated • Estimates suggest between 9% to

    47% of jobs can be automated • Not just factory workers • Japanese Company replaced insurance claim agents with IBM Watson
  9. Personalization • Machine Learning under our noses • Getting cheaper

    and easier • Product personalization popular, e.g. Nosto • Content personalization remains a challenge
  10. Fraud Detection • Used machine learning models for years •

    Accuracy will continue to increase • Become tailored to your business
  11. Conversational Commerce • New channels for Customer Service & Pre-Sales

    • Being available to customers at the right time
  12. Amazon Alexa • Estimated 5.2m sold last year • Customer

    spending increased 10% after purchase • Immersed in home life
  13. Natural Language Search • “Red waterproof jacket under $200” •

    Going beyond keywords to pull out and filter intent • Klevu are one of the leading examples of ML in search
  14. Chat Bots • Facebook Messenger • Order Confirmation and Updates

    • Automate answers to basic support requests
  15. Advanced Segmentation • Segmenting based on product history is nice

    • Segmenting customers based on their personality and desires • Tailor the messaging to them
  16. Intelligent Remarketing • We segment to avoid blasting all customers

    • Most ML budget is focused on predicting ad clicks • Reduce remarketing wastage
  17. Predictive Infrastructure Scaling • Scale before load is an issue

    • Learn from tell-tale signs • Spin up further resources so that they’ll be ready in time
  18. Anticipatory shipping • Patented by Amazon • Shipping to a

    general area based on predicted orders • Put the final address on package during transit
  19. Anomaly Detection • Static measurements of traffic or actions aren’t

    useful • Learn typical site usage taking into account • Seasonality • Time of day • Marketing campaigns
  20. Catalog Management • Detect Inconsistencies • “Shouldn’t this product be

    in the Handbags category?” • “Cameras normally have a 360 zoom image, should this 
 product have one?”
  21. Catalog Management • Imagine only having to categorize products once

    • New products automatically assigned to categories • Perhaps pre-fill attributes based on description
  22. What • Step back and profile your teams daily tasks

    • Experiment with time tracking to review where time goes • Discuss processes • Identify repetitive tasks • Include a technical representative if possible
  23. What • Busyness is dangerous to your business • It

    makes us feel productive and satisfied • Where is the value really coming from? • What are we wasting time on?
  24. How • Depends on your technical resources • If you

    have some, look into APIs, e.g. IBM Watson
  25. IBM Watson • Natural language Analysis • Solr Search that

    can be trained • Personality Insights • Tone Analyser • Trade-off analytics • Image recognition
  26. How • Supplier selection process • Choose tools that have

    incorporated ML • ML doesn’t make a product • But all great products will be using ML